src/librustc_ast/util/comments.rs

   1 pub use CommentStyle::*;
   2
   3 use crate::ast;
   4 use rustc_span::source_map::SourceMap;
   5 use rustc_span::{BytePos, CharPos, FileName, Pos};
   6
   7 use log::debug;
   8
   9 #[cfg(test)]
  10 mod tests;
  11
  12 #[derive(Clone, Copy, PartialEq, Debug)]
  13 pub enum CommentStyle {
  14     /// No code on either side of each line of the comment
  15     Isolated,
  16     /// Code exists to the left of the comment
  17     Trailing,
  18     /// Code before /* foo */ and after the comment
  19     Mixed,
  20     /// Just a manual blank line "\n\n", for layout
  21     BlankLine,
  22 }
  23
  24 #[derive(Clone)]
  25 pub struct Comment {
  26     pub style: CommentStyle,
  27     pub lines: Vec<String>,
  28     pub pos: BytePos,
  29 }
  30
  31 pub fn is_line_doc_comment(s: &str) -> bool {
  32     let res = (s.starts_with("///") && *s.as_bytes().get(3).unwrap_or(&b' ') != b'/')
  33         || s.starts_with("//!");
  34     debug!("is {:?} a doc comment? {}", s, res);
  35     res
  36 }
  37
  38 pub fn is_block_doc_comment(s: &str) -> bool {
  39     // Prevent `/**/` from being parsed as a doc comment
  40     let res = ((s.starts_with("/**") && *s.as_bytes().get(3).unwrap_or(&b' ') != b'*')
  41         || s.starts_with("/*!"))
  42         && s.len() >= 5;
  43     debug!("is {:?} a doc comment? {}", s, res);
  44     res
  45 }
  46
  47 // FIXME(#64197): Try to privatize this again.
  48 pub fn is_doc_comment(s: &str) -> bool {
  49     (s.starts_with("///") && is_line_doc_comment(s))
  50         || s.starts_with("//!")
  51         || (s.starts_with("/**") && is_block_doc_comment(s))
  52         || s.starts_with("/*!")
  53 }
  54
  55 pub fn doc_comment_style(comment: &str) -> ast::AttrStyle {
  56     assert!(is_doc_comment(comment));
  57     if comment.starts_with("//!") || comment.starts_with("/*!") {
  58         ast::AttrStyle::Inner
  59     } else {
  60         ast::AttrStyle::Outer
  61     }
  62 }
  63
  64 pub fn strip_doc_comment_decoration(comment: &str) -> String {
  65     /// remove whitespace-only lines from the start/end of lines
  66     fn vertical_trim(lines: Vec<String>) -> Vec<String> {
  67         let mut i = 0;
  68         let mut j = lines.len();
  69         // first line of all-stars should be omitted
  70         if !lines.is_empty() && lines[0].chars().all(|c| c == '*') {
  71             i += 1;
  72         }
  73
  74         while i < j && lines[i].trim().is_empty() {
  75             i += 1;
  76         }
  77         // like the first, a last line of all stars should be omitted
  78         if j > i && lines[j - 1].chars().skip(1).all(|c| c == '*') {
  79             j -= 1;
  80         }
  81
  82         while j > i && lines[j - 1].trim().is_empty() {
  83             j -= 1;
  84         }
  85
  86         lines[i..j].to_vec()
  87     }
  88
  89     /// remove a "[ \t]*\*" block from each line, if possible
  90     fn horizontal_trim(lines: Vec<String>) -> Vec<String> {
  91         let mut i = usize::MAX;
  92         let mut can_trim = true;
  93         let mut first = true;
  94
  95         for line in &lines {
  96             for (j, c) in line.chars().enumerate() {
  97                 if j > i || !"* \t".contains(c) {
  98                     can_trim = false;
  99                     break;
 100                 }
 101                 if c == '*' {
 102                     if first {
 103                         i = j;
 104                         first = false;
 105                     } else if i != j {
 106                         can_trim = false;
 107                     }
 108                     break;
 109                 }
 110             }
 111             if i >= line.len() {
 112                 can_trim = false;
 113             }
 114             if !can_trim {
 115                 break;
 116             }
 117         }
 118
 119         if can_trim {
 120             lines.iter().map(|line| (&line[i + 1..line.len()]).to_string()).collect()
 121         } else {
 122             lines
 123         }
 124     }
 125
 126     // one-line comments lose their prefix
 127     const ONELINERS: &[&str] = &["///!", "///", "//!", "//"];
 128
 129     for prefix in ONELINERS {
 130         if comment.starts_with(*prefix) {
 131             return (&comment[prefix.len()..]).to_string();
 132         }
 133     }
 134
 135     if comment.starts_with("/*") {
 136         let lines =
 137             comment[3..comment.len() - 2].lines().map(|s| s.to_string()).collect::<Vec<String>>();
 138
 139         let lines = vertical_trim(lines);
 140         let lines = horizontal_trim(lines);
 141
 142         return lines.join("\n");
 143     }
 144
 145     panic!("not a doc-comment: {}", comment);
 146 }
 147
 148 /// Returns `None` if the first `col` chars of `s` contain a non-whitespace char.
 149 /// Otherwise returns `Some(k)` where `k` is first char offset after that leading
 150 /// whitespace. Note that `k` may be outside bounds of `s`.
 151 fn all_whitespace(s: &str, col: CharPos) -> Option<usize> {
 152     let mut idx = 0;
 153     for (i, ch) in s.char_indices().take(col.to_usize()) {
 154         if !ch.is_whitespace() {
 155             return None;
 156         }
 157         idx = i + ch.len_utf8();
 158     }
 159     Some(idx)
 160 }
 161
 162 fn trim_whitespace_prefix(s: &str, col: CharPos) -> &str {
 163     let len = s.len();
 164     match all_whitespace(&s, col) {
 165         Some(col) => {
 166             if col < len {
 167                 &s[col..]
 168             } else {
 169                 ""
 170             }
 171         }
 172         None => s,
 173     }
 174 }
 175
 176 fn split_block_comment_into_lines(text: &str, col: CharPos) -> Vec<String> {
 177     let mut res: Vec<String> = vec![];
 178     let mut lines = text.lines();
 179     // just push the first line
 180     res.extend(lines.next().map(|it| it.to_string()));
 181     // for other lines, strip common whitespace prefix
 182     for line in lines {
 183         res.push(trim_whitespace_prefix(line, col).to_string())
 184     }
 185     res
 186 }
 187
 188 // it appears this function is called only from pprust... that's
 189 // probably not a good thing.
 190 pub fn gather_comments(sm: &SourceMap, path: FileName, src: String) -> Vec<Comment> {
 191     let sm = SourceMap::new(sm.path_mapping().clone());
 192     let source_file = sm.new_source_file(path, src);
 193     let text = (*source_file.src.as_ref().unwrap()).clone();
 194
 195     let text: &str = text.as_str();
 196     let start_bpos = source_file.start_pos;
 197     let mut pos = 0;
 198     let mut comments: Vec<Comment> = Vec::new();
 199     let mut code_to_the_left = false;
 200
 201     if let Some(shebang_len) = rustc_lexer::strip_shebang(text) {
 202         comments.push(Comment {
 203             style: Isolated,
 204             lines: vec![text[..shebang_len].to_string()],
 205             pos: start_bpos,
 206         });
 207         pos += shebang_len;
 208     }
 209
 210     for token in rustc_lexer::tokenize(&text[pos..]) {
 211         let token_text = &text[pos..pos + token.len];
 212         match token.kind {
 213             rustc_lexer::TokenKind::Whitespace => {
 214                 if let Some(mut idx) = token_text.find('\n') {
 215                     code_to_the_left = false;
 216                     while let Some(next_newline) = &token_text[idx + 1..].find('\n') {
 217                         idx = idx + 1 + next_newline;
 218                         comments.push(Comment {
 219                             style: BlankLine,
 220                             lines: vec![],
 221                             pos: start_bpos + BytePos((pos + idx) as u32),
 222                         });
 223                     }
 224                 }
 225             }
 226             rustc_lexer::TokenKind::BlockComment { terminated: _ } => {
 227                 if !is_block_doc_comment(token_text) {
 228                     let code_to_the_right = match text[pos + token.len..].chars().next() {
 229                         Some('\r' | '\n') => false,
 230                         _ => true,
 231                     };
 232                     let style = match (code_to_the_left, code_to_the_right) {
 233                         (_, true) => Mixed,
 234                         (false, false) => Isolated,
 235                         (true, false) => Trailing,
 236                     };
 237
 238                     // Count the number of chars since the start of the line by rescanning.
 239                     let pos_in_file = start_bpos + BytePos(pos as u32);
 240                     let line_begin_in_file = source_file.line_begin_pos(pos_in_file);
 241                     let line_begin_pos = (line_begin_in_file - start_bpos).to_usize();
 242                     let col = CharPos(text[line_begin_pos..pos].chars().count());
 243
 244                     let lines = split_block_comment_into_lines(token_text, col);
 245                     comments.push(Comment { style, lines, pos: pos_in_file })
 246                 }
 247             }
 248             rustc_lexer::TokenKind::LineComment => {
 249                 if !is_doc_comment(token_text) {
 250                     comments.push(Comment {
 251                         style: if code_to_the_left { Trailing } else { Isolated },
 252                         lines: vec![token_text.to_string()],
 253                         pos: start_bpos + BytePos(pos as u32),
 254                     })
 255                 }
 256             }
 257             _ => {
 258                 code_to_the_left = true;
 259             }
 260         }
 261         pos += token.len;
 262     }
 263
 264     comments
 265 }